Informs Annual Meeting 2017

TC67

INFORMS Houston – 2017

TC67

1 - A Regression Tree Method to Analyze Data in Functional Brain Activation Experiments Doowon Choi, doowon.choi@tamu.edu Functional brain activation experiments are widely conducted in neuroimaging research. In the experiment, behavior and functional brain activation of people under certain task are measured in an attempt to understand the underlying neural correlates in human brain. This work proposes a regression tree based methodology to analyze data in such experiments. In the case study, this method is applied to data from a functional near-infrared spectroscopy (fNIRS) study using a risk decision-making task and insightful inference is obtained. 2 - A Dictionary Learning Based Contemporaneous Health Index for Degenerative Disease Monitoring Aven Samareh, University of Washington, Seattle, WA, 98105, United States, asamareh@uw.edu, Shuai Huang We present a general formulation for personalized data fusion of multivariate longitudinal clinical measurements to create a contemporary health index (CHI) to monitor patient condition over time. Named as DL-CHI, our model combines longitudinal data fusion, classification, and dictionary learning and is applied on two real-world problems, Alzheimer’s disease and surgical site infection. 3 - Optimal Coverage Control in Body Sensor Networks for Spatiotemporal Cardiac Monitoring Rui Zhu, Pennsylvania State University, University Park, PA, United States, rzz45@psu.edu, Bing Yao, Hui Yang Body Sensor Networks (BSNs) has emerged as a key technology for improving the quality of life through wearable sensing, smart health monitoring. However, a fundamental problem in BSN design is the optimal sensor placement. For example, sensors are approximately uniformly distributed on the body surface in current ECG systems. Little has been done to optimize the location and number of sensors. This paper proposes a new strategy for the optimal sensor placement to capture a complete picture of cardiac dynamics. Furthermore, this novel methodology is applicable to network design with nonstationary distribution in many other disciplines, e.g., environment surveillance network, facility layout. 4 - Osteonecrosis in Chilean Patients with Systemic Lupus Erythematosus Jennifer Mendoza-Alonzo, University of South Florida, Tampa, FL, United States, jennifermend@mail.usf.edu, José Zayas-Castro, Karina Soto-Sandoval Systemic Lupus Erythematosus (SLE) is chronic, incurable, and autoimmune disease. A secondary disease associated with SLE is Osteonecrosis (ON). The objective of this study is to develop a model to determine if a Chilean patient who has already SLE can be diagnosed with ON at an early stage. The sample size is 41 Chilean patients with SLE. The input variables are pharmacological, demographic, and risk factors suggested by the literature. The outcome variable is the development of the disease ON. With this study is expected to contribute with more information about the multiple-factors that influence the development of ON to allow early detection of this diseases in patients with SLE. 371D Spatial Optimization and Conservation Reserve Design Sponsored: Energy, Natural Res & the Environment Environment & Sustainability Sponsored Session Chair: Bistra Dilkina, Georgia Institute of Technology, Atlanta, GA, 30332, United States, bdilkina@gmail.com 1 - A Robust Optimization Approach for Solving Problems in Conservation Planning Zulqarnain Haider, University of South Florida, ENG 302, University of South Florida, Industrial and Management Systems Engineering, Tampa, FL, 33620, United States, zulqarnain@mail.usf.edu, Hadi Charkhgard, Changhyun Kwon, Julien Martin In conservation planning, the data related to specie populations is sparse, and unreliable. Thus, applying deterministic or stochastic approaches to the problems in conservation planning either ignores the uncertainty completely or does not accurately account for the nature of uncertainty. We propose a robust optimization (RO) approach to problems in conservation planning. We explore the reserve selection and invasion control problems to show the value of the RO approach. Several novel techniques are developed to compare the results by the RO approach and the deterministic approach. It is demonstrated that the proposed approach finds more applicable conservation planning strategies. TC69

371B Data-driven Approaches to Predictive Analytics Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Abdallah A. Chehade, University of Michigan-Dearborn, 815 Eagle Heights Apt D, Madison, WI, 53705, United States, abdallahch89@gmail.com Co-Chair: Raed Al Kontar, University of Wisconsin-Madison, Madison, WI, 53706, United States, alkontar@wisc.edu 1 - A Multi-task Learning Approach for Improved Forecast of Passenger Inflow Time Series Public transportation services such as, Urban Railway Transit (URT) systems, play an increasingly important role in people’s commute. However, there has been little research on the influence of passenger crowding on the URT systems. To fill this research gap, we propose a state space model to predict the passenger inflow time series under the multi-task learning framework. By incorporating engineering knowledge into the prior, the methodology would produce more accurate and robust estimate of the parameters. We adopt the EM algorithm and Laplace’s method for MAP estimate of the parameters. A real case study on Hong Kong’s MTR passenger flows is presented to illustrate the effectiveness of our model. 2 - System-wide Prediction and Visualization of Passenger Distribution in Urban Rail Transit Network with Automatic Fare Collection Data Andi Wang, Georgia Institute of Technology, Atlanta, GA, United States, andi.wang@connect.ust.hk, Min Jiang, Fugee Tsung We propose a scheme to achieve real-time passenger distribution prediction in the urban metro system, based on the entry and exit time stamps generated by the automatic fare collection system. First, we infer each individual’s travel pattern and location distribution with historical data, and store the information in the data center. Then the information is integrated with real-time entry and exit data streams to achieve passenger distribution prediction with acceptable computational cost. The effectiveness of the scheme is validated with real data set from the MTR system in Hong Kong. 3 - Functional Data Based Computer Model Calibration for Design Optimization Wenbo Sun, University of Michigan, 2013 Medford Road, Ann Arbor, MI, 48104, United States, sunwbgt@umich.edu, Judy Jin, Matthew Plumlee Computer simulations are widely used to improve system design. This often requires to firstly calibrate the computer simulation model by using a UQ (uncertainty quantification) method and then search for an optimal design solution. The existing methods require explicitly defined simulation parameters (e.g. to minimize the simulation model’s bias) and a well fitted metamodel that can accurately surrogate the simulation behaviors, which are hard for complex systems. We presented a new approach to search for the optimal design from a set of parameters that subject to constraints on simulation model’s bias. A case study is illustrated for the vehicle design optimization to minimize injuries in crash. 4 - Structural Degradation Modeling Framework for Sparse Data Sets with an Application on Alzheimer’s Disease Abdallah A. Chehade, University of Michigan-Dearborn, Dearborn, MI, United States, abdallahch89@gmail.com, Kaibo Liu The accessibility and development of information technologies have facilitated the collection and storage of information from a massive number of operating units, which provides a great opportunity to better understand the degradation process. While in practice the number of recorded units can be large, there are often limited available observations in most of the units. To address the unique challenges in the “sparse data environments”, this paper proposes a structural degradation modeling (SDM) framework. On one hand, our idea is inspired by the recommender system. On the other hand, the proposed SDM is also tailored to the needs of degradation modeling. 371C Quality and Statistical Decision Making in Health Care Applications Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Shuai Huang, University of Washington, Seattle, WA, 98195, United States, shuai.huang.ie@gmail.com TC68 Zhenli Song, HKUST, Hong Kong, Hong Kong, zsongae@connect.ust.hk, Ke Zhang, Fugee Tsung

360